Specific Armadillo Repeat Sequences Facilitate β-Catenin Nuclear Transport in Live Cells via Direct Binding to Nucleoporins Nup62, Nup153, and RanBP2/Nup358*

Background: The nuclear localization of β-catenin is directly linked to its cancer causing activity. Results: Armadillo repeats (10–12) mediate nuclear transport of β-catenin through direct interaction with specific nuclear pore complex proteins. Conclusion: β-Catenin can function like a nuclear transport receptor in its ability to translocate independently through the nuclear pore complex. Significance: β-Catenin may transport specific binding partners between the nucleus and cytoplasm in response to Wnt signaling. β-Catenin transduces the Wnt signal from the membrane to nucleus, and certain gene mutations trigger its nuclear accumulation leading to cell transformation and cancer. β-Catenin shuttles between the nucleus and cytoplasm independent of classical Ran/transport receptor pathways, and this movement was previously hypothesized to involve the central Armadillo (Arm) domain. Fluorescence recovery after photobleaching (FRAP) assays were used to delineate functional transport regions of the Arm domain in living cells. The strongest nuclear import/export activity was mapped to Arm repeats R10–12 using both in vivo FRAP and in vitro export assays. By comparison, Arm repeats R3–8 of β-catenin were highly active for nuclear import but displayed a comparatively weak export activity. We show for the first time using purified components that specific Arm sequences of β-catenin interact directly in vitro with the FG repeats of the nuclear pore complex (NPC) components Nup62, Nup98, and Nup153, indicating an independent ability of β-catenin to traverse the NPC. Moreover, a proteomics screen identified RanBP2/Nup358 as a binding partner of Arm R10–12, and β-catenin was confirmed to interact with endogenous and ectopic forms of Nup358. We further demonstrate that knock-down of endogenous Nup358 and Nup62 impeded the rate of nuclear import/export of β-catenin to a greater extent than that of importin-β. The Arm R10–12 sequence facilitated transport even when β-catenin was bound to the Arm-binding partner LEF-1, and its activity was stimulated by phosphorylation at Tyr-654. These findings provide functional evidence that the Arm domain contributes to regulated β-catenin transport through direct interaction with the NPC.

␤-Catenin transduces the Wnt signal from the membrane to nucleus, and certain gene mutations trigger its nuclear accumulation leading to cell transformation and cancer. ␤-Catenin shuttles between the nucleus and cytoplasm independent of classical Ran/transport receptor pathways, and this movement was previously hypothesized to involve the central Armadillo (Arm) domain. Fluorescence recovery after photobleaching (FRAP) assays were used to delineate functional transport regions of the Arm domain in living cells. The strongest nuclear import/export activity was mapped to Arm repeats R10 -12 using both in vivo FRAP and in vitro export assays. By comparison, Arm repeats R3-8 of ␤-catenin were highly active for nuclear import but displayed a comparatively weak export activity. We show for the first time using purified components that specific Arm sequences of ␤-catenin interact directly in vitro with the FG repeats of the nuclear pore complex (NPC) components Nup62, Nup98, and Nup153, indicating an independent ability of ␤-catenin to traverse the NPC. Moreover, a proteomics screen identified RanBP2/Nup358 as a binding partner of Arm R10 -12, and ␤-catenin was confirmed to interact with endogenous and ectopic forms of Nup358. We further demonstrate that knock-down of endogenous Nup358 and Nup62 impeded the rate of nuclear import/export of ␤-catenin to a greater extent than that of importin-␤. The Arm R10 -12 sequence facilitated transport even when ␤-catenin was bound to the Arm-binding partner LEF-1, and its activity was stimulated by phosphorylation at Tyr-654. These findings provide functional evidence that the Arm domain contributes to regulated ␤-catenin transport through direct interaction with the NPC.
The canonical Wnt signaling pathway regulates normal tissue development and maintenance through its effector molecule, ␤-catenin. In the absence of Wnt signaling ␤-catenin accumulates at adherens junctions. ␤-Catenin levels are regulated through degradation by a destruction complex, comprising of adenomatous polyposis coli (APC), 4 axin, casein kinase I, and glycogen synthase kinase-3␤, which phosphorylates and subsequently ubiquitinates ␤-catenin at the N terminus, leading to its degradation by the proteosome (1,2). On receiving Wnt signals or through misregulation of the degradation pathway, ␤-catenin is stabilized and accumulates in the nucleus where it binds to lymphoid enhancer factor-1 (LEF/ TCF) transcription factors and activates transcription of tumor promoting genes involved in cell migration, invasion, and proliferation (3). Thus, nuclear accumulation of ␤-catenin is directly linked to cell transformation. Deregulation of this pathway is implicated in a number of cancers including colon, breast, liver, and skin (3)(4)(5). Furthermore, immunohistochemical studies identified increased nuclear accumulation of ␤-catenin in cells at the invasive front of differentiated mesenchyme-like tumors (6 -8), and in colorectal cancers of increasing tumor grade (9). Therefore, ␤-catenin nuclear accumulation is linked to cancer.
␤-Catenin is a unique 90-kDa protein that shuttles in and out of the nucleus but does not possess a classical nuclear export signal or nuclear localization signal required for transportation via exportin/importin pathways, and does not require Ran-GTPase for transport (10 -12). We and others observed that a fraction of ␤-catenin can exit the nucleus indirectly in complex with other proteins (e.g. APC, Kank, LZTS2, Axin) that do access the CRM1/exportin-1 route, at least when these proteins are overexpressed in cells (13)(14)(15)(16)(17). However, when its expression is induced transiently by Wnt signaling or chronically by cancer-linked mutations, the majority of ␤-catenin exits the nucleus independent of CRM1, exogenous soluble factors, and Ran-GTPase (12,18). Additionally, the nuclear import of ␤-catenin occurs independently of Ran-GTPase and the importins (10,11), although LEF-1 has been implicated in its import via the importin pathway (19). Notably, the receptor independent pathway for nuclear transport of ␤-catenin has not yet been resolved.
Structurally, ␤-catenin comprises a helical folded 12 Armadillo (Arm) repeat sequence flanked by unstructured N and C termini (20,21) (see Fig. 1A). It was previously proposed (and frequently assumed in the literature) that the ␤-catenin Arm domain mediates nuclear transport, as Arm repeats 9 -12 adopt a very similar conformation to importin-␤ HEAT repeats 4 -8, which mediate nuclear transit through direct contact with the NPC (22,23). However, there is currently little evidence to support a role for the Arm repeat domain in ␤-catenin nuclear transport or for interaction with NPC components. First, the complete Arm repeat sequence (R1-12), or a combination of the Arm and C terminus (R8 -12 ϩ C), displayed only very weak transport activity in a Xenopus oocyte microinjection assay (18) or in photobleaching assays in human cells (24). Moreover, a study by Koike et al. (22) could not measure any transport activity of the Arm sequence alone, and it was claimed that only in combination with C-terminal sequences did Arm repeats R10 -12 contribute to transport of ␤-catenin using in vitro digitonin cell permeabilization assays and microinjection of cells. In terms of evidence for binding to FG repeat containing Nups, Fagotto et al. (10) showed in vitro that ␤-catenin could bind directly to the FG repeats of a single yeast nucleoporin, Nup1p, however, they did not assess the FG repeats of mammalian Nups normally contacted by transport receptors. Moreover, the same laboratory later rescinded their claims and reported that ␤-catenin does not bind to Nup FG repeats (25). More recently, Hendriksen et al. (26) cited unpublished data that the Arm domain of ␤-catenin could immunoprecipitate certain nucleoporins from Xenopus laevis oocytes, but no testing for a direct interaction between ␤-catenin and NPC components was performed.
In this report, we tested different ␤-catenin Arm sequences and show for the first time that the Arm domain is highly active for transport, and that this activity maps primarily to Arm repeats 10 -12. We further demonstrate differential transport activity of other Arm repeat sequences and show evidence for a direct interaction between ␤-catenin Arm sequences and the FG repeats of specific nucleoporins (Nups) implicated in protein transport. The live cell photobleaching assay data are further compounded by evidence that silencing specific Nups slows transport of ␤-catenin. These findings outline a model for the nuclear transport of ␤-catenin.

Cell Culture, RNA Interference, and Transfections
NIH 3T3 cells were routinely maintained in 75-cm 2 tissue culture flasks by passaging with trypsin/EDTA. At 24 h prior to transfection, cells were seeded on coverslips in a 6-well tray or 2-well chamber slides (Nunc) and transfected at ϳ50% confluence using Lipofectamine according to the manufacturer's instructions. Cells were treated with 5.2 ng/ml of leptomycin B (LMB) for 3 h prior to fluorescence recovery after photobleaching (FRAP). Mouse Nup358 (sc-36381) and Nup62 siRNA (sc-36108) sequences were ordered from Santa Cruz. NIH 3T3 cells were seeded at 50% confluence for 24 h prior to transfection. 3 g of siRNA was transfected using Lipofectamine (Invitrogen) according to the manufacturer's protocol. Cells were given fresh media after 5 h of transfection. After 48 h these cells were transfected with ␤-catenin-WT-GFP or ␤-catenin fragments using FuGENE (Promega) according to the manufacturer's protocol. 72 h post-transfection, cells were imaged (for FRAP assay), immunostained, or lysed for Western blot assay.

Live Cell FRAP Assay
FRAP analysis was performed on an Olympus FV1000 confocal laser scanning microscope using ϫ60 water objective; zoom ϭ 2, scanning speed ϭ fast, image format ϭ 512 ϫ 512, laser power during pre-and post-bleach scanning ϭ ϳ5-7% to minimize loss of fluorescence intensity, and laser power during bleaching ϭ 100%. The microscope chamber was set and equilibrated to 37°C and 5% CO 2 .
Nuclear Export/Import FRAP Data Acquisition-3 prebleach images of the whole cell were acquired. 90% of the cytoplasm/nucleus for export/import FRAP was bleached for 9 -12 s (depending on the construct and stability of GFP tag in these cells). Post-bleach imaging was done in 3 stages: first stage, 30 frames at the fastest interval; second stage, 30 frames at 1.2-s intervals; third stage, 30 frames at 10-s intervals. Fluorescence intensities for the cytoplasm, nucleus, and background were acquired using Olympus Fluoview software and exported to a Microsoft Excel file.
Data Analysis-Background values were subtracted from cytoplasmic and nuclear fluorescence intensities. To compare the rates of nuclear export/import between different samples, fluorescence data were expressed as a cytoplasmic/nuclear ratio (for nuclear export) or nuclear/cytoplasmic ratio (for nuclear import). For each cell data set, the pre-bleach ratio was set to 100%, the time for the first post-bleach image was set to 0 s, and the recovery curve was adjusted to start at 25% recovery at time 0 (this was the closest average value). The average of the data for at least 10 cells from 2 to 4 experimental repeats was plotted. Initial export/import rates for the first 30 s was analyzed in GraphPad Prism 5.0 using linear regression analysis. Other kinetic analysis was done using the single association curve in GraphPad Prism 5.0 with the curve weighted by 1/Y 2 .

Immunoprecipitation
Cells were cross-linked by incubating the flask with 0.75 mM dithiobis(succinimidylpropionate)/PBS as described by the supplier (Pierce) for 20 min at room temperature. The reaction was stopped by incubating the flask with 50 mM glycine/TBS for 5 min at room temperature. Cells were scraped and lysed using RIPA buffer (27). For immunoprecipitation analysis 2 mg of protein lysate was incubated with 1 g of the specified antibody overnight at 4°C with continuous end-over-end mixing. The immunocomplexes were collected with 30 l of protein G-agarose (Amersham Biosciences) (for mouse monoclonal antibody) or protein A-Sepharose (for rabbit anti-GFP antibody) beads by incubation at 4°C for 2 h with continuous end-overend mixing. The immunocomplexes were pulled down by centrifugation and washed 3 times with TBS prior to being denatured in Laemmli buffer. The immunoprecipitates were subjected to SDS-PAGE followed by immunoblotting.

Western Blotting
NIH 3T3 cells were seeded in a T25 flask and transfected with ␤-catenin-GFP fragments using Lipofectamine (as described in supplemental Methods). 48 h post-transfection cells were harvested using trypsin and lysed in RIPA buffer for 20 min on ice (containing fresh protease inhibitor mixture from Roche Diagnostics). Lysates were spun at 13,000 ϫ g for 10 min at 4°C. The supernatant was quantified using a Bradford assay. 50 g of total cell lysate was separated on a 7.5% SDS-PAGE gel and transferred onto a nitrocellulose membrane. The membrane was blocked with 5% skim milk/PBS and immunoblotted with anti-GFP antibody (1:1000 from Roche Diagnostics) and antimouse HRP antibody (1:5,000 from Sigma).

In Vitro Binding Assay
MBP fusions of ␤-catenin were expressed and purified from DH5␣ bacteria, and glutathione S-transferase (GST)-tagged FG repeats of Nup62, -98, and -153 were expressed and purified from BL21 cells (for details of protein purification refer to supplemental Methods). 10 l of amylose resin (New England Biolabs) was coated with ϳ500 ng of purified MBP or ␤-catenin-MBP for 1 h at 4°C, washed once with TEDM buffer (TEDM ϭ 20 mM Tris, pH 8.0, 1 mM EDTA, 1 mM DTT, 1 mM MgCl 2 ), and blocked with 5% BSA/TEDM for 1 h at 4°C. 500 ng of purified GST or Nup-GST fragments were added to blocked amylose resin and incubated for 2 h at 4°C. The resin was washed once with RIPA buffer and twice with TEDM ϩ 0.25 M NaCl. The resin was eluted with 10 l of maltose elution buffer and Laemmli buffer. The eluate was loaded and separated by SDS-PAGE and then transferred onto nitrocellulose membrane. The membrane was first probed with anti-GST antibody and then with anti-MBP antibody. Similar exposures were quantified with Adobe Photoshop. ␤-Catenin-WT-MBP and MBP were always included as controls and for quantification purpose. For comparison, background signal from MBP protein was subtracted and the ratio of ␤-catenin-WT-MBP to Nup-GST bound to the beads was set to 100%. Each experiment was repeated 2-4 times.

Mass Spectrometry Approach
Immunoprecipitates were separated by SDS-PAGE, stained with Coomassie Blue, and appropriate bands were excised. Proteins were digested with trypsin, peptides were recovered and analyzed by liquid chromatography/mass spectrometry (LC/ MS) using an LTQ XL linear ion trap (ThermoScientific) as described (34). To identify peptides the spectral data were searched using MASCOT with tolerances of 1.2 and 0.6 Da for precursor and fragment masses, respectively.

Statistical Analysis
Statistical analysis was performed on the data using the STATVIEW program (version 5; SAS Institute). Drug treatment and transfection comparisons were analyzed by one-way analysis of variance. Statistics for FRAP assays were calculated using the repeated measure of analysis (one-way analysis of variance) over multiple time points and post hoc t tests were used to compare significant differences between constructs. Results were considered significant when p Ͻ 0.05. The Student's unpaired t test was also used.

RESULTS
The Arm Repeats 10 -12 of ␤-Catenin Display Strong Nuclear Export Activity in Living Cells-It was previously speculated that specific Arm repeats (9 -12) of ␤-catenin (Fig. 1A) contribute to its nuclear transport due to a structural resemblance to the transport-active HEAT repeats of importin-␤ (22,23). However, published analyses of the ␤-catenin Arm region have provided no clear answer due to contradictory data (18,22,24). To compare the activity of different Arm regions of ␤-catenin, we constructed GFP-tagged Arm repeat fragments for use in FRAP assays in NIH 3T3 cells. This noninvasive technique enabled us to compare both nuclear export and import rates of ␤-catenin sequences in living cells. The GFP-tagged peptides were all beyond the size limit (ϳ45 kDa) for passive diffusion through the NPC and correct expression of each construct was verified by Western blotting (supplemental Fig. S1).
␤-Catenin can exit the nucleus via the CRM1 pathway when bound to specific partners that contain an nuclear export sequence (13)(14)(15)(16)(17). However, the major export route of ␤-catenin is independent of CRM1 (35). To ensure that only CRM1-independent export was analyzed, cells were treated with the CRM1 inhibitor LMB for 3 h prior to photobleaching. To measure nuclear export, the GFP-based fluorescence of transfected cells was bleached in the cytoplasm and the rate of fluorescence recovery was then quantified for up to 360 s ( Fig.  1B and supplemental Fig. S2). For ease of comparison of transport rates, the different fluorescence recovery curves were plotted and shown as the cytoplasmic:nuclear (C/N) ratio (see "Materials and Methods") for the first 150 s (Fig. 1C). We confirmed our previous result showing the rapid CRM1-independent nuclear export of wild-type (WT) ␤-catenin relative to a 58-kDa tomato-dimer fluorescent protein control (36). The export rates of all the ␤-catenin subfragments were then assessed from recovery slopes by linear regression analysis of the first 30 s (Fig. 1D), during which time transport was least influenced by nuclear retention or re-equilibration. A comparison of the FRAP data revealed quite clearly that the Arm repeat (R1-12)-GFP fragment was exported equally as fast (if not slightly faster) as wild-type ␤-catenin ( Fig. 1, B-D). Surprisingly, this is the first report to demonstrate strong nuclear export activity of the central Arm repeats.
The rate of CRM1-independent export for ␤-catenin (fulllength or Arm R1-12) was found to be comparable with that of YFP-CRM1 in live cells (Fig. 1). It is interesting to note that LMB treatment actually caused a modest increase in ␤-catenin transport activity relative to untreated cells (supplemental Fig.  S2), suggesting that active CRM1 might normally compete with ␤-catenin for transport through the nuclear pore complex.
We further fine-mapped the transport activity, and the individual Arm repeat sequences R1-8, R3-8, and R10 -12 of ␤-catenin were expressed as GFP fusions and tested for nuclear export by FRAP (Fig. 1, B and C). The Arm sequences R1-8 and R3-8 were least efficiently exported, and recovered in the cytoplasm only up to 50% of pre-bleach values (Fig. 1C). In contrast, fluorescence levels of Arm sequence R10 -12 recovered rapidly up to ϳ80% of the pre-bleach cytoplasmic intensity, and displayed a similar recovery curve to that seen for the full Arm R1-12 sequence. A comparison of the initial export rates (summarized in Fig. 1D) distinguished Arm repeats R10 -12 as the primary export sequence. To underscore the specificity of this activity, the export kinetics of the 7-repeat Arm domain from APC protein was tested and found to be relatively slow for nuclear export and comparable in activity to ␤-catenin Arm repeats R1-8 and R3-8 (Fig. 1, B and C). This is the first systematic analysis of the ␤-catenin and APC Arm repeats in live cells, and reveals both differential export activity of different Arm sequences and direct evidence that Arm sequence R10 -12 comprises the strongest nuclear export function. The nuclear export activity of the R10 -12 was further confirmed using an FIGURE 1. Arm repeats (R10 -12) of ␤-catenin mediate nuclear export. A, a schematic diagram of the primary ␤-catenin protein domains: N terminus (1-141 amino acids), Arm repeats 1-12 (141-664 amino acids), and C terminus (665-781 amino acids). B, NIH 3T3 cells were transfected with ␤-catenin-GFP fragments (WT, R1-12, R1-8, R3-8, R10 -12, APC arm), YFP-CRM1 or td-tomato (58-kDa negative control) as indicated. After 48 h, cells were pretreated with 5.2 ng/ml of LMB for 3 h before nuclear export FRAP analysis. Confocal images are shown of cells before bleach and after Ͼ90% of the cytoplasm was bleached (shown up to 160s when nearing plateau). Side panel shows relative export activity. C, mean recovery curves for all constructs were calculated as the cytoplasmic to nuclear ratio, set to 100% based on prebleach values and compared with td-tomato (negative control). Each curve represents an average of at least 10 cells from 2 to 4 experiments. D, nuclear export rate (mean Ϯ S.D.) was calculated for the first 30 s using one-phase association equation and linear regression analysis on GraphPad Prism software. See supplemental Fig. S2 and Table S2 for a more detailed kinetic analysis.
␤-Catenin Arm R10 -12 Mediates Rapid Bidirectional Transport, Whereas ␤-Catenin Arm R3-8 Is More Biased toward Nuclear Import Than Export-A previous study by Krieghoff and colleagues (24) used FRAP analysis to study ␤-catenin transport, but did not distinguish between nuclear import and export rates and did not compare different Arm repeat sequences. We next employed FRAP analysis to test the nuclear import kinetics of the GFP-tagged Arm repeats by bleaching the nuclear fluorescence with a high-power laser and tracking the rate of nuclear recovery over 150 s. Note that all transfected cells analyzed displayed an equivalent and moderate level of nuclear and cytoplasmic GFP expression. Arm sequences R1-12 and R10 -12 were found to be capable of rapid nuclear import at a rate slightly faster than wild-type ␤-catenin (Fig. 2, A-C). Interestingly, the import rates of ␤-catenin wild-type and Arm (R1-12 or R10 -12) sequences were ϳ40 -50% slower than that of importin-␤, a benchmark import receptor. This suggests that importin-␤ is slightly more active at the NPC, and indeed a previous study indicated that importin-␤ could effectively compete with ␤-catenin for transport across the nuclear pores (22). Further analysis revealed that ␤-catenin Arm sequence R3-8 was also imported into the nucleus relatively efficiently (Fig. 2, A and B), and at a faster rate than it was exported (compare graphs in Figs. 1D and 2C, and see supplemental Fig. S4). These findings pinpoint ␤-catenin Arm region R10 -12 as the optimal bidirectional import/export element, and identify R3-8 as a strong import sequence with limited export activity. Our results distinguish for the first time the potential for differential transport activity (i.e. import versus export) of different Arm repeat sequences in living cells (see supplemental Table S2 for details).
The Arm Repeats of ␤-Catenin Bind Directly to Transport Active FG Repeats of Nucleoporins-The receptor-independent transport of ␤-catenin is reminiscent of transport receptors such as importin-␤ and NTF2 that traverse the NPC by binding to multiple FG repeats of nucleoporins such as Nup62, Nup98, and Nup153 (37). An interaction of wild-type ␤-catenin with yeast Nup1p was originally reported by Fagotto et al. (10) but later contradicted by the same group who could not confirm this interaction (25). To address this issue, we first performed an immunoprecipitation experiment, using an antibody highly specific for nucleoporin FG repeats to pull down binding partners in cross-linked SW480 colon tumor cell extracts. As shown in Fig. 3A, the immunoprecipitation assay detected a weak in vivo interaction with ␤-catenin, suggesting a physiological interaction between ␤-catenin and Nups at the NPC.
Next, we tested whether the transport active Arm repeats of ␤-catenin identified here by the FRAP assay could bind directly in vitro to specific NPC regions thought to mediate active passage of transport receptors, i.e. the FG repeats of Nup62, Nup98, and Nup153. ␤-Catenin sequences (wild-type, R3-8, and R10 -12) were purified from bacteria as MBP fusions and coated onto amylose resin beads. GST-tagged forms of the Nup FG repeats were then prepared and compared for binding to the MBP-␤-catenin sequences (protocol outlined in Fig. 3B). There was very little background binding of GST-Nups to beads coated with MBP alone (Fig. 3C, first and fourth lanes). Wildtype ␤-catenin-MBP was found to bind consistently above background to the FG repeats of Nup62, -98, and -153 (e.g. second lane). In Fig. 3D, the percentage binding of MBP-␤catenin sequences to FG repeats was calculated by normalizing against MBP, and setting wild-type ␤-catenin-MBP values to 100% (see "Materials and Methods"). The Arm repeats of ␤-catenin were bound to Nup FG repeats, although to varying FIGURE 2. Arm repeats (R10 -12) of ␤-catenin mediate rapid nuclear import. A, NIH 3T3 cells were transfected with different ␤-catenin Arm-GFP plasmids or GFP-importin-␤ (as indicated) and at 48 h post-transfection cells were subjected to nuclear import FRAP analysis. Confocal cell images are shown before bleaching and after the nucleus was bleached using 100% laser power; fluorescence recovery was monitored over 400 s (shown up to 160 s when near or at plateau). Side panel shows the relative import activity. B, mean recovery curves were calculated for each construct as the nuclear to cytoplasmic ratio, setting pre-bleach fluorescence values to 100% and compared with td-tomato. Each curve represents at least 10 cells from 2 to 4 experiments. C, nuclear import rate (mean Ϯ S.D.) was calculated for the first 30 s using a one-phase association equation and linear regression analysis on GraphPad Prism software. See supplemental Table S2 for detailed kinetic analysis.
Arm-mediated ␤-Catenin Nuclear Transport extents. It is interesting to note that Arm repeats R3-8 showed differential binding to these Nups (bound Nup62 much more strongly than Nup153), which might possibly correlate with its slow rate of nuclear export relative to import (Fig. 2). These data provide the first clear demonstration, using purified components, that full-length ␤-catenin and its Arm repeat domain can bind directly to specific nucleoporin FG repeats. This is highly consistent with a direct transport route for ␤-catenin through the NPC.
The Arm Repeat Sequence R10 -12 Binds to FG Repeats of RanBP2/Nup358-The Arm repeat R10 -12 region displayed the strongest nuclear export and import activities (Figs. 1 and  2), and thus we used this sequence to screen for other potential binding partners that might influence its nuclear transit activ-ity. MBP-␤-catenin R10 -12-coated beads were incubated with total cell lysate from NIH 3T3 cells and candidate binding partners were identified by mass spectrometry (see supplemental Fig. S5). This approach identified several new binding partners, of which RanBP2/Nup358 was the most promising new transport partner. RanBP2/Nup358 is a mobile nucleoporin that locates at the cytoplasmic filaments of the NPC. We confirmed that ␤-catenin interacts with Nup358 by immunoprecipitation, where endogenous full-length ␤-catenin was pulled down by Nup358 antibody (see Fig. 3F). In addition, the FG repeat fragment of Nup358 was transiently expressed in cells and shown to specifically immunoprecipitate ␤-catenin (Fig. 3E). To compare ␤-catenin-Nup358 binding with that of transport receptors, we co-transfected the FG repeats of Nup358 with ␤-catenin-GFP, GFP-importin-␤, or YFP-CRM1 and immunoprecipitated using anti-GFP antibody. As anticipated, a modest and comparable interaction was observed between Nup358-FG repeats and the three transporter proteins (Fig. 3G). Taken together, these results suggest that ␤-catenin interacts with the FG repeats of Nup358 (cytoplasmic filaments), Nup62 (central channel), Nup98 and Nup153 (nuclear basket) and traverses the NPC without the aid of transport receptors (see model in Fig.  5H).
Silencing of Nup358 and Nup62 Slows Nuclear Export and Import of ␤-Catenin-No previous study has yet tested the functional influence of specific Nups on ␤-catenin transport. To evaluate whether Nup358 regulates ␤-catenin transport, we silenced its expression by siRNA in NIH 3T3 cells, achieving a selective loss of up to 90% (Fig. 4G and supplemental Fig. S6). We then studied the live cell transport kinetics of GFP-tagged ␤-catenin Arm R10 -12 in response to silencing Nup358. As shown in Fig. 4A, silencing Nup358 did not affect the steadystate nuclear-cytoplasmic distribution of ␤-catenin (see prebleach panel). However, in FRAP experiments, the silencing of Nup358 decreased nuclear export of ␤-catenin-R10 -12-GFP by up to 40% of the rate seen with control siRNA (see Fig. 4, B and C). Thus, Nup358 binds to the Arm R10 -12 sequence and contributes to its transport dynamics.
Next, we tested whether loss of either Nup358 or Nup62, which are located at the cytoplasmic face and central channel of the NPC, respectively, impacted the transport of full-length ␤-catenin. FRAP analysis of ␤-catenin-GFP expressed in NIH 3T3 cells revealed that relative to control siRNA, the knockdown of Nup358 caused a significant ϳ45% reduction in nuclear export rate (Fig. 4, D and E), and a ϳ50% reduction in import rate of ␤-catenin (Figs. 4, F and G, and 5E). A similar knockdown of mouse Nup62 was performed in NIH 3T3 cells and the effect on kinetics of ␤-catenin-GFP were evaluated. Nup62 was effectively silenced as verified by Western blot (Fig.  5G), and its loss had no effect on ␤-catenin steady-state distribution (see pre-bleach image in Fig. 5A). However, the Nup62 knockdown caused a ϳ30% reduction in export rate (Fig. 5, A,  B, and E) and a ϳ20% decrease in nuclear import rate (Fig. 5, C-E) of ␤-catenin. To compare with another transport receptor, we silenced Nup62 and Nup358 and studied the effect on nuclear import of GFP-importin-␤. We found that loss of neither nucleoporin had any significant effect on the transport rate of importin-␤ ( Fig. 5F and supplemental Fig. S7).
These findings suggest that both Nup62 and Nup358 contribute, with a degree of selectivity, to the trafficking of ␤-catenin through the NPC (Fig. 5H). We note in particular that loss of Nup358 had a more pronounced effect on ␤-catenin than did loss of Nup62, implicating an important role of the NPC cytoplasmic filaments in the transport process. The changes observed in the nuclear transport rate are quite significant and specific, in that the transport rate of ␤-catenin was shown previously not to be affected by blocks to Wnt signaling (38), or by inhibition of CRM1 activity (supplemental Fig. S2), and the knockdown of Nup62 or Nup358 did not alter general nuclear envelope integrity or the nuclear-cytoplasmic diffusion of GFP (supplemental Fig. S6). This is the first evidence to demonstrate that loss of specific nucleoporins affects nuclear shuttling of ␤-catenin.
The Arm R10 -12 Sequence Is Active for Export Even When ␤-Catenin Is Bound to LEF-1-The ability of ␤-catenin to translocate across the NPC like a transport receptor suggests that it might function as a selective bidirectional carrier of specific binding partners. Many of the primary partners that regulate ␤-catenin retention (e.g. LEF-1/TCFs and cadherins) bind within the first 8 Arm repeats, raising the question of whether Arm repeats R10 -12 would still be accessible and functional for nuclear transport when bound to such cargo in the presence of LMB. To test this, we used the LEF-1⌬HMG construct as a test cargo. This form of LEF-1 lacks the DNA-binding domain and therefore is not retained at the chromatin. However, LEF-1⌬HMG retains the nuclear localization signal (28) and therefore locates independently in the nucleus, differing only from wild-type LEF-1 in that it is free to be exported by a carrier (see FIGURE 3. Purified ␤-catenin binds directly to nucleoporin FG repeats. A, endogenous Nups from SW480 cells were immunoprecipitated (IP) using mAb414 antibody (recognizes FG repeats from Nup62 and -153) after dithiobis(succinimidylpropionate)-mediated cross-linking, separated by SDS-PAGE, and analyzed by Western blot. The membrane was probed with ␤-catenin monoclonal antibody or mAb414 antibody (to confirm the pull-down). IgG was also used for immunoprecipitation as a background control. B, protocol diagram for in vitro binding assay. Amylose beads were coated with purified MBP protein or ␤-catenin-MBP protein (␤) and blocked with 5% BSA/TEDM buffer. The coated beads were then incubated with purified GST protein or GST-tagged FG repeats of Nup62, -98, and -153. After extensive washing, bound protein was eluted with maltose elution buffer, boiled in sample buffer, separated by SDS-PAGE, and immunoblotted for detection with anti-GST or anti-MBP antibody. C, Western blot probed with anti-GST and anti-MBP antibodies is shown from a representative experiment. Beads coated with MBP (negative control) and wild-type (WT) ␤-catenin-MBP (positive control) proteins were included in each pulldown experiment. D, the percentage of GST-Nups bound to ␤-catenin-MBP protein was quantified using bands from similar exposures from 3 different experiments. For comparison, signal from ␤-catenin-WT-MBP protein was set to 100% (see "Materials and Methods"). Statistically significant differences from wild-type are indicated. E, a plasmid encoding HA-tagged FG repeats of Nup358 was co-transfected with ␤-catenin-WT-GFP into NIH 3T3 cells. Cell lysates were subjected to immunoprecipitation using ␤-catenin mAb or isotype control IgG antibody. The blot was probed with anti-HA antibody or ␤-catenin mAb and shows that ␤-catenin interacts with FG repeats of Nup358. F, cell lysates from SW480 cells were immunoprecipitated using Nup358 mAb or isotype control IgG antibody. The blot was probed with ␤-catenin mAb or Nup358 mAb and shows that these full-length endogenous proteins interact in the cell. G, pNup358-FG-HA was co-transfected with pYFP-CRM1, pGFP-importin-␤, or p␤-catenin-GFP in NIH 3T3 cells and cells were then immunoprecipitated with anti-GFP antibody. The blot was probed with anti-HA, revealing interaction between transport receptors and FG repeats of Nup358. Fig. 6A). This form of LEF-1 binds ␤-catenin at Arm repeats R3-8, which leaves R10 -12 potentially available for export (Fig. 6). We co-transfected LEF-1⌬HMG with different ␤-catenin-GFP sequences (Fig. 6B) and scored the localization of LEF-1⌬HMG in co-transfected cells by microscopy. We found that wild-type ␤-catenin-GFP shifted staining of LEF-1⌬HMG from the nucleus to cytoplasm in 85% of cells, compared with just 20% of cells co-transfected with GFP alone. The deletion of the N-terminal tail, or both the N-and C-terminal tails, did not diminish export activity of ␤-catenin-GFP, with the expression of the full Arm domain (R1-12) alone shifting LEF-1⌬HMG to the cytoplasm in ϳ90% of cells. In contrast, transient expression of the Arm R10 -12 sequence, which does not bind to LEF-1, displayed a similar background activity to GFP control (Fig. 6B). This data suggests that Arm repeats R10 -12 can contribute to transport of cargo out of the nucleus even when proteins are bound to the adjacent Arm repeat sequences of ␤-catenin.
Tyrosine Mutations in Arm R10 -12 Sequence (Tyr-654) Show Differential Transport Activity-We recently showed that the S45A mutation, which blocks glycogen synthase kinase-3␤ phosphorylation of the N terminus of ␤-catenin, has no significant effect on nuclear transport (38). Src-mediated phosphorylation of tyrosine 654, which lies within the Arm R10 -12 FIGURE 4. Silencing Nup358 impedes nuclear export and import of ␤-catenin. A, NIH 3T3 cells were transfected with control siRNA or Nup358 siRNA and co-transfected with ␤-catenin-(R10 -12)-GFP for nuclear export FRAP analysis. Confocal cell images are shown before bleaching and after the cytoplasm was bleached using 100% laser power; fluorescence recovery was monitored over 300 s (images shown up to 160 s). B, mean recovery curves of ␤-catenin-(R10 -12)-GFP in the presence of control siRNA or Nup358 siRNA for up to 300 s. Each curve represents at least 10 cells from 2 to 3 experiments. C, nuclear export rates (mean Ϯ S.D.) were calculated for the first 30 s from the initial recovery slopes using linear regression analysis. D, a similar FRAP assay for nuclear export was performed on NIH 3T3 cells transfected with wild-type (WT) ␤-catenin-GFP, and typical cell images are shown of cells before and after photobleaching the cytoplasm. E, export recovery curves are shown for cells transfected with control or Nup358 siRNA. F, FRAP analysis of nuclear import was performed on ␤-catenin-GFP-transfected cells, and cell images are shown. G, import recovery curves displaying the nuclear/cytoplasmic ratio over time. H, Western blot analysis to confirm knock-down of Nup358.

Arm-mediated ␤-Catenin Nuclear Transport
sequence (see Fig. 7A), has been shown to dissociate ␤-catenin from E-cadherin at the plasma membrane (30), and to stimulate Wnt signaling and initiate APC-dependent tumorigenesis in mice (39). We tested whether mutagenesis of this residue affects ␤-catenin nuclear transport, by performing live cell FRAP analysis of mutants Y654E (phosphomimic) and Y654F (nonphosphorylated) in full-length ␤-catenin-GFP. Unexpectedly, both mutations caused a ϳ50% increase in the import rate of ␤-catenin, and displayed comparable T1 ⁄ 2 values (Fig. 7, B, C, and F). When analyzed for export, however, Y654E showed a similar nuclear export rate to that of wild-type ␤-catenin-GFP, whereas Y654F was significantly slower (Fig. 7, D, E, and G). We conclude that perturbation of a single residue in the active R10 -12 sequence is sufficient to modulate ␤-catenin transport, and that shifts in phosphorylation of Tyr-654 can impact on nuclear export dynamics.

DISCUSSION
␤-Catenin can shuttle between three distinct subcellular locations: the nucleus, cytoplasm, and plasma membrane. In live cells it was previously shown by photobleaching assays that the Arm domain, a known protein interaction site, contributes to retention of ␤-catenin in the nucleus and at cell:cell junctions through binding to LEF-1/TCF transcription factors and cad-FIGURE 5. Silencing Nup62 slows nuclear transport of ␤-catenin. A, NIH 3T3 cells were transfected with control siRNA or Nup62 siRNA and co-transfected with wild-type ␤-catenin-GFP for nuclear export FRAP analysis. Confocal cell images are shown before bleaching and after the cytoplasm was bleached using 100% laser power; fluorescence recovery was monitored over 360 s. B, mean export recovery curves of ␤-catenin-GFP in the presence of control siRNA or Nup62 siRNA. Each curve represents at least 10 cells from 2 to 3 experiments. C, FRAP analysis was performed on ␤-catenin-GFP transfected into NIH 3T3 cells as above, but analyzing nuclear import rate. Cell images are shown. D, recovery curves for nuclear import are shown. Nuclear export and import rates (Ϯ S.D.) were calculated for (E) ␤-catenin-GFP and (F) GFP-importin-␤ during the first 30 s and compared for control, Nup358 and Nup62 siRNA-treated cells. G, Western blot analysis to confirm knock-down of Nup62. H, schematic model for receptor-independent transport of ␤-catenin (␤) through the NPC. ␤-Catenin transiently binds to the FG repeats of Nup358 (located at the cytoplasmic filaments), Nup62 (central channel of NPC), Nup98 and Nup153 (located at the nuclear basket of NPC).

Arm-mediated ␤-Catenin Nuclear Transport
herins, respectively (24,36,38). Here, we have re-assessed the potential role of the Arm repeat domain to mediate active transport of ␤-catenin. We show for the first time in live cells that the 12-repeat Arm region retains the same degree of nuclear import/export activity as full-length ␤-catenin, and we mapped the optimal bidirectional transport function to Arm repeats R10 -12. The R10 -12 sequence was found to bind to the FG repeats of Nup98 and Nup153 (located on the nuclear face of the NPC), Nup62 (central channel), and Nup358 (cytoplasmic face of the NPC). Thus, ␤-catenin uses the same transport route as nuclear transport receptors, and displayed similar transport kinetics as importin-␤ and CRM1. In functional assays, it was further revealed that silencing of Nup62 or -358, which are located at different positions within the NPC, diminished the import/export rate of ␤-catenin. This provides the first evidence supporting the conjecture that the ␤-catenin Arm R9 -12 region might contact the NPC, based on its folding similarities to the importin-␤ HEAT repeats (23). These new lines of evidence, when considered together, strongly support a role for the Arm domain in contributing to ␤-catenin nuclear transit through direct and transient contacts with the NPC, and resolves some of the discrepancies in the literature that likely arose from different methodologies applied to measure transport and Nup binding (10,25).
Our results, furthermore, elucidate differences between different Arm sequences, exemplified by ␤-catenin Arm repeats R3-8 and the APC Arm domain, which displayed a significantly slower export rate than ␤-catenin Arm R10 -12. The differential transport activity (import Ͼ export) of the Arm R3-8 sequence correlated with a difference in its binding to specific Nups, showing a pronounced preference for Nup62 relative to Nup153. This may suggest that passage and direc-FIGURE 6. The Arm R10 -12 sequence is export-active when ␤-catenin is bound to LEF-1. A, representative cell images from fluorescence microscopy of NIH 3T3 cells transfected with GFP alone or different ␤-catenin-GFP fragments, and co-transfected with pLEF-⌬HMG. The left panel indicates the ␤-catenin fragments used, and the binding site for LEF-1 (Arm repeats R3-8) relative to the position of Arm repeats R10 -12. Right-hand panel indicates cells with nuclear or cytoplasmic localization of LEF-⌬HMG, after cells were treated 3 h with 5.2 ng/ml of LMB, fixed, then scored by microscopy for protein distribution at 48 h post-transfection. The Arm domain alone was capable of shifting LEF-1 from the nucleus to cytoplasm. B, cells were scored and data are plotted to show % cells with partial or complete cytoplasmic localization of LEF1-⌬HMG after co-expression of different ␤-catenin fragments. Ͼ200 co-transfected cells were scored from 2 to 4 experiments. tionality of the Arm R3-8 sequence across the NPC is affected by an affinity gradient, as previously reported for importin-␤ (40). The various Arm repeats vary substantially in sequence identity but adopt a typical helical fold (21); the differences between R10 -12 and R3-8 could thus arise from the exposure of specific amino acids at the helical surface. Indeed, NPC-contacting domains of transport receptors are generally presumed to be surface-located hydrophobic patches and are dependent on protein conformation (reviewed in Ref. 41). In this context we note that changes to a single residue (Tyr-654) of the R10 -12 sequence were sufficient to accelerate ␤-catenin import to a rate comparable with that of importin-␤ (Fig. 7). Conversely, a nonphosphorylatable Y654F mutation significantly decreased the export rate of ␤-catenin. This highlights the potential for certain post-translational modifications to impact on the ␤-catenin Wnt signaling function through modulation of transport efficacy.
It is interesting to note that silencing Nup62 and Nup358/ RanBP2 did not affect the steady state localization of ␤-catenin-GFP (pre-bleach panels in Figs. 4 and 5), but did impact on the kinetics of nuclear transport. In particular, despite effective silencing of both Nups, the loss of Nup358 caused a far more consistent disruption of ␤-catenin import/export than loss of Nup62. At the NPC, Nup62 exists in a complex with Nup45, Nup54, and Nup58 in the central channel (42,43) and also interacts with Nup153, Nup214, and Nup358 (44). Therefore, the loss of Nup62 may be more readily compensated by other Nups. The silencing of Nup62 was recently reported to slow the nuclear export of HIV Rev viral RNA particles, although actual transport dynamics were not measured (45).
Nup358 is the major component of cytoplasmic filaments and its roles in docking/undocking of transport substrates is less readily compensated for by other Nups. This may explain why knockdown of Nup358 elicited a stronger effect on trans-FIGURE 7. Tyr-654 mutations in full-length ␤-catenin display differential transport activity. A, diagram of ␤-catenin showing sites of phosphorylation by glycogen synthase kinase-3␤ (GSK-3␤) and c-Src and for ubiquitination by ␤-TrCP. B, NIH 3T3 cells were transfected with ␤-catenin-GFP Y654E or Y654F (as indicated) for nuclear import FRAP analysis. Confocal cell images are shown before bleaching and after the nucleus was bleached using 100% laser power; fluorescence recovery was monitored over 250 s. C, mean import recovery curves for Y654E and Y654F. Each curve represents at least 10 cells from 2 to 3 experiments. D, NIH 3T3 cells were transfected with ␤-catenin-GFP Y654E or Y654F for nuclear export FRAP in the presence of LMB. Confocal cell images are shown before and after bleaching. E, mean export recovery curves Y654E or Y654F. Each curve represents at least 10 cells from 2 to 3 experiments. F and G, nuclear import and export rates (mean Ϯ S.D.) for Y654E and Y654F mutants was compared with wild-type ␤-catenin-GFP. port of ␤-catenin in this study. Studies employing both in vitro (46,47) and in vivo (48) assays have shown that siRNAmediated loss of Nup358 can cause a reduction in importin-␣/␤ and transportin-mediated nuclear import of proteins such as HIV Rev. In addition, recent studies found that Nup358 promotes nuclear import of proteins in a selective transport receptor-specific manner (49), which partly involves the capture and recycling of RanGTP⅐importin-␤ complexes at the cytoplasmic fibrils of the NPC (50). We observed no change in import rate for total GFP⅐importin-␤ after knockdown of Nup62 or Nup358 (supplemental Fig. S7), suggesting that only its Ran-dependent transfer of nuclear localization signal-containing cargo is dependent on Nup358 docking. We propose that ␤-catenin binds transiently to sequential Nups during its passage through the NPC. Because the knockdown of Nup358 caused a selective 50% decrease in nuclear export/import rate of ␤-catenin, we speculate that Nup358 is important for ␤-catenin docking/undocking at NPC cytoplasmic fibrils during the nuclear transport process.
In a previous study the Arm repeats of ␤-catenin were implicated in a RanGTP-dependent export pathway. It was proposed that the CRM1 export co-factor, RanBP3, binds to the Arm repeats of ␤-catenin in a RanGTP-stimulated fashion and exports the transcriptionally active, de-phosphorylated pool of ␤-catenin out of the nucleus independent of the CRM1 export pathway (26). This resulted in reduced transcriptional activity of ␤-catenin and reduced effects of Wnt signaling in X. laevis and Drosophila melanogaster. Although this paper was intriguing, we could not replicate the findings and observed no changes in translocation of de-phosphorylated ␤-catenin under similar conditions (i.e. in LiCl-treated NIH 3T3 cells or SW480 cells) after overexpression or silencing of RanBP3 (see details under supplemental Fig. S8). It is possible some discrepancies arise from the antibody originally used to recognize dephospho-␤-catenin, which we (supplemental Fig. S8A) and others have found to be nonspecific (51).
The Arm domain is the main protein interaction region of ␤-catenin, and it is targeted by a range of key binding and/or retention partners including LEF-1/TCFs, APC, and cadherins (52). The majority of these factors bind within the first 8 repeats, and thus are likely to mask the Arm R3-8 sequence, but it was not obvious if the Arm R10 -12 would remain functional. To address this question we co-transfected cells with various ␤-catenin sequences and a nuclear nonretained mutant of LEF-1 (Fig. 6), and found that the full-length Arm could indeed function to export nuclear-localized LEF-1 to the cytoplasm. Thus, R10 -12 remains accessible and active even when ␤-catenin is bound to its Arm-binding cargo. This has implications for the role of ␤-catenin as a potential transporter of various cargo (19,53). A key distinction between ␤-catenin and other transport receptors is that ␤-catenin is not regulated by Ran-GTPase and thus, rather than dissociate from its cargo after crossing the NPC, it is likely to move with its protein partners along extended transport routes between the nuclear chromatin and the plasma membrane. In summary, we propose that Arm R10 -12 contributes to shuttling of ␤-catenin between the nucleus and cytoplasm by directly contacting FG repeats of critical nucleoporins, and that when stabilized by Wnt signaling or cancer mutations, ␤-catenin is likely to utilize this transport pathway and function as a protein-specific nuclear transport receptor.     Tables and Figures S1-S8